Agricultural and Forest Meteorology 100 2000 199–212
Estimating tree crown dimensions using digital analysis of vertical photographs
P.L. Brown
a,b,∗
, D. Doley
a,b
, R.J. Keenan
a,c
a
Cooperative Research Centre for Tropical Rainforest Ecology and Management, Cairns, Queensland 4870, Australia
b
Department of Botany, The University of Queensland, St Lucia, Queensland 4072, Australia
c
Bureau of Rural Sciences, P.O. Box E11, Kingston, ACT 2604, Australia Received 10 May 1999; received in revised form 30 August 1999; accepted 17 September 1999
Abstract
The aim of this study was to develop a relatively rapid, simple and repeatable method to characterize the crown architecture of Queensland maple Flindersia brayleyana F. Muell. by vertical hemispherical photography. Calibration scales were
developed to provide horizontal planes of measurement at heights from 15 to 40 m above the camera. Hemispherical images of tree crowns were merged with the specific calibration scale appropriate for the mid-height of the crown in order to adjust
for image distortion during measurement. Merged images were analyzed by three procedures to yield measurements of crown diameter and projected area. The most precise measurement of projected crown dimensions was assumed to be provided by
a complete analysis of a digital image of the crown, after removal of the trunk image. Crown diameters estimated by tape measurements on the ground were relatively imprecise R
2
= 0.54. Crowns classified as symmetrical were described much
more precisely R
2
= 0.72 than those classified as asymmetrical R
2
= 0.37. The precision of estimates of crown dimensions
by tape did not alter significantly over the range of crown sizes sampled 2–10 m diameter in trees 15–40 m tall. This novel image analysis method can yield relatively precise and accurate measurements on the relatively shallow crowns of tall trees,
often found in rainforests, regardless of the degree of asymmetry in crown shapes. However, the method is slower to apply than conventional tape techniques and additional equipment is required. The principal advantage of the photographic technique is
the opportunity for precise measurements of changes in crown size and foliage projected area over time. ©2000 Published by Elsevier Science B.V. All rights reserved.
Keywords: Tree crowns; Crown asymmetry; Crown diameter; Crown projected area; Hemispherical photography; Radial distortion
1. Introduction
Vertical photographs of forest canopies provide detailed information about radiation regimes within
forests, which can be used to predict individual tree and stand growth Koop, 1989. In tropical rainforests,
∗
Corresponding author. Tel.: +61-7-3365-2768; fax: +61-7-3365-1699.
E-mail address: p.brownbotany.uq.edu.au P.L. Brown.
where the availability of light is most limiting to the growth of individual trees below the main canopy,
applications of radiation measurements may prove crucial for improving existing tree growth prediction
models Landsberg and Gower, 1997. Vertical pho- tographs currently contain some of the most precise
and detailed information on forest radiation regimes but their utility has been restricted largely by the
difficulty of making accurate repeatable estimates. Application of radiation-based tree growth models
0168-192300 – see front matter ©2000 Published by Elsevier Science B.V. All rights reserved. PII: S 0 1 6 8 - 1 9 2 3 9 9 0 0 1 3 8 - 0
200 P.L. Brown et al. Agricultural and Forest Meteorology 100 2000 199–212
to individual trees requires improvements in the de- scription of radiation interactions by their crowns in
structurally complex forests. These interactions need to be estimated with considerable precision, and im-
portant small-scale changes in crown dimensions are difficult to estimate in any tall or complex forests,
such as those of the wet tropics.
There is increasing global interest in expanding the range of species used in timber plantations, as
a substitute for native forest utilization. This inter- est is particularly strong in the wet tropics, where
many broad-leaved species are being proposed for plantation management Evans, 1982; Rowe et al.,
1991; Ferris-Kaan, 1992. The successful adoption of broad-leaved species, many with irregular crown
shapes at a relatively early age, depends largely on the definition of their optimum silvicultural requirements.
An understanding of a species crown architecture is fundamental to designing silvicultural prescriptions
Dawkins, 1963; Suri, 1975; Samarasinghe, 1995; Dhanesh Kumar et al., 1997, and this understanding
is usually built upon accurate measurements of plant dimensions which reflect the photosynthetic potential
of individual trees.
Plant dimensions such as crown projected area and diameter are often used as surrogates for crown
photosynthetic potential and foliage area Jahnke and Lawrence, 1965, but they are usually derived from
relatively imprecise measurements taken from the ground Philip, 1994. Consequently, there is a need
for better techniques for measuring these attributes.
The measurement of crown dimensions can be par- ticularly difficult in rainforest broad-leaved species,
because of irregular crown shapes. Light environments of tropical rainforests are often more heterogeneous
than those of temperate forests Whitmore and Burn- ham, 1984 and the development of asymmetric crown
shapes has been postulated as a plastic response to this heterogeneity Umeki, 1995. Most conventional
methods for measuring crown diameter and projected area assume that crown projections are either a solid
circle or solid ellipsoid shape Wilkinson, 1995; Big- ing and Gill, 1997. Therefore, such methods vary
in precision according to the degree crowns depart from simple geometric shapes Biging and Gill, 1997;
Zeide, 1998.
Airborne and satellite remote sensing techniques, that measure specific forest attributes, are rapidly im-
proving in their discriminative capacity. For example, surface Lidar remote sensing systems now have the
capacity to rapidly obtain information at the stand level on canopy height, total biomass, leaf biomass,
basal area and percentage cover Means et al., 1999. Lidar systems can operate at a nominal footprint
diameter of 10 m yet currently lack the capacity to allow measurements of the dimensions of individ-
ual tree crowns Lefsky et al., 1999. Stereo pairs of co-registered aerial photographs have been used
effectively to monitor changes in co-dominant and emergent tropical rainforest crown dimensions over
the longer term 18 years Herwitz et al., 1998a. However, even when using photographs taken at
large-scales from low flying aircraft, such techniques currently lack the discrimination needed to accurately
monitor smaller changes in tree crown dimensions in complex forest systems, particularly on individuals
with irregular crown shapes or those growing in sub- dominant positions Herwitz et al., 1998b. Vertical
photography has also been used to determine tree crown dimensions. Koop 1989 used stereo-pairs
of vertical photographs taken a short distance apart to convert small image displacements between the
photographs into distance and crown area estimates. Acceptable precision was achieved even on smaller
crown sizes, but at considerable cost per tree. Koike 1985 reconstructed two-dimensional canopy pro-
files from clusters of vertical photographs processed by computed tomography. This procedure is very de-
manding of resources particularly when the distance between the camera and subject varies substantially.
Chen et al. 1991 measured plant area index using both the analysis of hemispherical photographs and
a Plant Canopy Analyzer Li-Cor LAI-200 on the same crowns. They found that both methods were in
agreement when the hemispherical photographs were underexposed four to five stops below that recom-
mended by an upward facing light meter. Although, the study by Chen et al. 1991 refines and validates
the use of hemispherical photos to measure plant area index the need remains for a relatively quick, sim-
ple and accurate method of estimating small crown diameters on tall trees.
An example of a tropical forest species which fre- quently has small, irregular shaped crowns on tall trees
and for which crown architectural descriptions are needed is Queensland maple Flindersia brayleyana F.
P.L. Brown et al. Agricultural and Forest Meteorology 100 2000 199–212 201
Muell.. This species is highly prized as a cabinet tim- ber and is endemic to the wet tropics region of eastern
Australia, where its supply from native forests has all but ceased Cameron and Jermyn, 1991. Bazzaz and
Pickett 1980 suggest maple is a middle-order suc- cession species and evidence is increasing that it can
survive in a wide range of light conditions Thompson
Fig. 1. Images of crown projected areas typifying the asymmetrical group six above and symmetrical group six below. A circle of known dimensions at a known height encompasses each crown however the relative scale of these circles has been altered for individual
crowns to a standard diameter for the diagram.
et al., 1988; Swanborough et al., 1998. With weak apical control and highly plastic growth, maple tends
to develop asymmetric crowns that can be difficult to describe adequately by simple geometric shapes see
Fig. 1. This irregularity may be due to crown shyness Jacobs, 1955; Ng, 1976, with crown development
being strongly influenced by surrounding crowns.
202 P.L. Brown et al. Agricultural and Forest Meteorology 100 2000 199–212
The aim of this study was to develop a simple, re- peatable method for measuring crown diameter and
projected area using vertical photographs. This aim evolved out of the need to characterize the crown ar-
chitecture of Queensland maple, for we were partic- ularly interested in obtaining accurate measurements
on crowns with irregular or asymmetrical shapes.
2. Methodology
